up

app.py

@@ -1,14 +1,7 @@
-import requests
-import json
-import pandas as pd
-from tqdm.auto import tqdm
 import streamlit as st
 from pandas import read_csv
 import os
-from huggingface_hub import HfApi, hf_hub_download
-from huggingface_hub.repocard import metadata_load
 import jiwer
-import datetime
 from huggingface_hub import Repository
 
 REFERENCE_NAME = "references"
@@ -118,7 +111,11 @@ COLUMN_NAMES = {
 
 table = all_results
 
-tabel = table.select_dtypes(exclude=['object', 'string']) * 100
+esc_column = table.pop("esc-score")
+name_column = table.pop("name")
+table.insert(0, "esc-score", esc_column)
+table = table.select_dtypes(exclude=['object', 'string']) * 100
+table.insert(0, "name", name_column)
 table = table.round(2)
 table = table.rename(columns=COLUMN_NAMES)
 
@@ -128,20 +125,20 @@ st.markdown("# ESC: A Benchmark For Multi-Domain End-to-End Speech Recognition")
 st.markdown(
     f"""
     This is the leaderboard of the End-to end Speech Challenge (ESC).
-    Submitted systems are ranked by the **ESC Score** which is the average of 
+    Submitted systems are ranked by the **ESC Score** which is the average of
     all non-optional datasets: {', '.join(COLUMN_NAMES.values())}."""
 )
 
-#st.table(table)
+# st.table(table)
 table
 
 #    *Sanchit Gandhi, Patrick Von Platen, and, Alexander M Rush*
 st.markdown(
     """
     ESC was proposed in *ESC: A Benchmark For Multi-Domain End-to-End Speech Recognition* by ...
-    \n 
-    The abstract of the paper is as follows: 
-    \n 
+    \n
+    The abstract of the paper is as follows:
+    \n
     *Speech recognition applications cover a range of different audio and text distributions, with different speaking styles, background noise, transcription punctuation and character casing. However, many speech recognition systems require dataset-specific tuning (audio filtering, punctuation removal and normalisation of casing), therefore assuming a-priori knowledge of both the audio and text distributions. This tuning requirement can lead to systems failing to generalise to other datasets and domains. To promote the development of multi-domain speech systems, we introduce the End-to end Speech Challenge (ESC) for evaluating the performance of a single automatic speech recognition (ASR) system across a broad set of speech datasets. Benchmarked systems must use the same data pre- and post-processing algorithm across datasets - assuming the audio and text data distributions are a-priori unknown. We compare a series of state-of-the-art (SoTA) end-to-end (E2E) systems on this benchmark, demonstrating how a single speechsystem can be applied and evaluated on a wide range of data distributions. We find E2E systems to be effective across datasets: in a fair comparison, E2E systems achieve within 2.6% of SoTA systems tuned to a specific dataset. Our analysis reveals that transcription artefacts, such as punctuation and casing, pose difficulties for ASR systems and should be included in evaluation. We believe E2E benchmarking over a range of datasets promotes the research of multi-domain speech recognition systems.*
     \n
     For more information, please see the official submission on [OpenReview.net](https://openreview.net/forum?id=9OL2fIfDLK).